CN106570192A - Deep learning-based multi-view image retrieval method - Google Patents
Deep learning-based multi-view image retrieval method Download PDFInfo
- Publication number
- CN106570192A CN106570192A CN201611014355.1A CN201611014355A CN106570192A CN 106570192 A CN106570192 A CN 106570192A CN 201611014355 A CN201611014355 A CN 201611014355A CN 106570192 A CN106570192 A CN 106570192A
- Authority
- CN
- China
- Prior art keywords
- image
- view
- view image
- deep learning
- optimum
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/50—Information retrieval; Database structures therefor; File system structures therefor of still image data
- G06F16/58—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually
- G06F16/583—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually using metadata automatically derived from the content
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Library & Information Science (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Databases & Information Systems (AREA)
- Artificial Intelligence (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Image Analysis (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The invention provides a deep learning-based multi-view image retrieval method. A training process of the method comprises steps of normalizing a multi-view image from aspects of dimensionality and scale, building a multi-view deep convolution neural network, building a convolution neural network of multi-path view parallel processing, enabling initial parameters of each path of network to be the same, refining network parameters, refining and adjusting pre-trained network parameters via a label data set, classifying images in an image base and conducting optimization view angle calculation after the training, extracting an image feature and an optimal view angle of each group of multi-view, and storing the image feature and the optimal view angle. During image retrieval, after a single view is input, a feedback result via retrieval comprises a similar image and an image having the optimal view angle. By the use of the method of retrieving the multi-view via the single view, retrieval accuracy and result displaying intuition can be improved.
Description
Technical field
The present invention relates to a kind of field of image search, especially a kind of multi-view image retrieval side based on deep learning
Method.
Technical background
Traditional image retrieval mode is to enter line retrieval using single-view.Figure to the different visual angles of same destination object
As constituting one group of image sets for being better able to pictute destination object, this image sets are referred to as multi-view image.More than these
In view image, some views can accurately embody destination object, but some visible images do not have expressive ability so,
The visual angle for being best able to characterize in these multi-view images multi-view image target is referred to as optimum visual angle.Multi-view image application is very
Extensively, such as it is typically with multi views to embody in the online displaying of article of e-commerce platform;Design patent image
It is that one kind represents product appearance image using multi views.
Multi-view image is not come by traditional search method to scheme to search figure as organic whole, causes retrieval accurate
Rate is relatively low.Meanwhile, when retrieval source images are selected, what the result of retrieval was often fed back is the image similar to source images,
Sometimes source images are deposited in itself and can not extremely accurate identify target itself, therefore the result for causing to retrieve is unsatisfactory.Cause
This, it is need to solve in technical field one to design a kind of method that can retrieve optimum multi-view image in multi-view image
Problem.
The content of the invention
For deficiency of the prior art, the present invention provides a kind of multi-view image search method based on deep learning,
Multi views convolutional neural networks are built and train, the network parameter by the use of pre-training is adopted as the initial weight of learning network
The weight of labelling multi-view image regularized learning algorithm network, after completing study, by image zooming-out feature and optimum in data base
Visual angle is simultaneously stored in data base, and image is input in learning network during retrieval the feature for extracting image and its optimum visual angle,
Then it is compared with the characteristics of image in data base, returns the optimal view image of similar image.The method is substantially increased
The accuracy of image searching result, solves feature association and optimal characteristics table between multi-view image retrieving multi-view image
The problem shown.
According to design provided by the present invention, a kind of multi-view image search method based on deep learning, its instruction
Practice process specifically to comprise the steps of:
Step 1. multi-view image pretreatment, by multi-view image dimension normalization, image dimension normalization, while will be many
Each view of view image makes a distinction classification;Multi-view image data collection is divided into into test data set and training dataset two
Part.
Step 2. constructs multi views depth convolutional neural networks, is that each class view adopts VGG-M nets according to view classification
Network parameter, simply is exported using newSoftmax to replace by Softmax classification by rear, is made using the network parameter of pre-training
For the initial weight of network.
Step 3. network parameter becomes more meticulous adjustment, reversely adjusts network parameter with training dataset, further with band
The test data set of label adjusts network parameter, the multi views deep learning network model after being trained becoming more meticulous.
In step 4. data base, characteristics of image and optimum visual angle calculate, the spy of the multi-view image in calculated off line data base
Levy and its optimum visual angle, and result is stored in property data base.
It is above-mentioned, in step 1, graphical rule normalization be by Image Adjusting be identical yardstick.
Preferably, yardstick is 227*227.
Above-mentioned, in step 1, image dimension normalization is that the gray level image of two dimension is changed into three-dimensional similar rgb format
Image.
Preferably, R, G, the value of channel B respective pixel and the gray level image respective pixel value of new image will be increased
It is identical.
It is above-mentioned, in step 2, multi views branch be respectively front view, left view, right view, top view, upward view, after
View and three-dimensional view branch.
Above-mentioned, in step 2, the network parameter of pre-training uses the network parameter for arriving trained based on VGG-M.
Preferably, VGG-M models are selected as network parameter.
Above-mentioned, in step 2, the initial parameter of multi views network branches is identical with the network architecture.
Above-mentioned, in step 2, vector of the result that newSoftmax is exported for M*N, wherein M represent the species of view, N tables
The classification number of diagram picture.
Above-mentioned, in step 4, multi-view image obtains the eigenvalue and each view of image by deep learning network calculations
Optimum visual angle.
According to design provided by the present invention, a kind of multi-view image search method based on deep learning, its inspection
Rope process is specifically comprised the steps of:
Step one. Image semantic classification, graphical rule normalization that will be to be retrieved, image dimension normalization.
Step 2. by image by any multi views depth convolutional neural networks all the way, it is calculated the feature of image.
Step 3. the feature of image is compared with the characteristics of image in data base, according to apart from output image from small to large
Call number, extracts the corresponding optimum multi-view image of call number from image library.
Step 4. retrieval result sequencing of similarity is exported, by the optimum multi-view image and similar image packet output of retrieval.
Beneficial effects of the present invention:The present invention is answered to optimal view retrieval for existing multi-view image retrieval shortage
With, using depth convolutional neural networks construct multi views convolutional neural networks, corresponding view is entered according to the path of view
Row process of convolution, analyzes the optimal view for being best able to represent multi views group, substantially increases the intuitive of image searching result
With the accuracy of image retrieval.
Description of the drawings
Fig. 1. the training schematic flow sheet of the present invention.
Fig. 2. the retrieval flow schematic diagram of the present invention.
Fig. 3. flow chart provided in an embodiment of the present invention.
Specific embodiment
In order that the purpose of the present invention, technical scheme are advantage become more apparent, it is below in conjunction with drawings and Examples, right
The present invention is further described.It should be appreciated that specific embodiment described herein is only to explain the present invention, and without
It is of the invention in limiting.
Embodiment one, with reference to shown in Fig. 1, a kind of multi-view image search method based on deep learning, it is characterised in that
Training process includes:
Step 1. multi-view image pretreatment, by multi-view image dimension normalization, image dimension normalization, while will be many
Each view of view image makes a distinction classification;Multi-view image data collection is divided into into test data set and training dataset two
Part.
Step 2. constructs multi views depth convolutional neural networks, is that each class view adopts VGG-M nets according to view classification
Network parameter, simply is exported using newSoftmax to replace by Softmax classification by rear, is made using the network parameter of pre-training
For the initial weight of network.
Step 3. network parameter becomes more meticulous adjustment, reversely adjusts network parameter with training dataset, further with band
The test data set of label adjusts network parameter, the multi views deep learning network model after being trained becoming more meticulous.
In step 4. data base, characteristics of image and optimum visual angle calculate, the spy of the multi-view image in calculated off line data base
Levy and its optimum visual angle, and result is stored in property data base.
Embodiment two:With reference to shown in Fig. 2, a kind of multi-view image search method based on deep learning, it is characterised in that
Retrieving includes:
Step 201. Image semantic classification, graphical rule normalization that will be to be retrieved, image dimension normalization.
Image by any multi views depth convolutional neural networks all the way, is calculated the feature of image by step 202..
The feature of image is compared by step 203. with the characteristics of image in data base, according to apart from output image from small to large
Call number, extracts the corresponding optimum multi-view image of call number from image library.
Step 204. retrieval result sequencing of similarity is exported, will be the optimum multi-view image and similar image packet of retrieval defeated
Go out.
Embodiment three:With reference to shown in Fig. 3, a kind of multi-view image search method based on deep learning, it is characterised in that
Image is changed into 227*227 sizes by step 301. graphical rule normalization.
Step 302. image dimension normalization, if image is RGB three-dimensional, keeps constant;If image is gray scale
Image or bianry image, then be changed into the 3-D view of similar RGB by the bidimensional that increases of image, newly increase dimension with it is original
Image is identical.
Each road view of image is matched by step 303. with corresponding convolutional neural networks.
Step 304.CNN, will be by CNN convolutional calculation characteristics of image.
Step 305.FC2, characteristics of image extract characteristics of image by FC2 fully-connected networks, and the feature after FC2 is extracted is
4096 dimensions.
Step 3051.Feature1, using FC2 after each road view characteristic vector pickup out as the spy of multi views
Levy the Part I of vector.
Step 306.FC3, characteristics of image extract characteristics of image by FC3 fully-connected networks, and the feature after FC3 is extracted is
4096 dimensions.
Feature1 and Feature2 vectors are weighted draw and obtain characterizing multi views by step 3061.Feature2
The characteristic vector of image, and store.
Preferably, storage is compressed to image feature vector.
Step 3062. characteristics of image, using FC2 after each road view characteristic vector pickup out as the spy of multi views
Levy the Part I of vector.
Step 307.newSoftmax, feature of the image by newSoftmax network extraction images, forms the spy of M*N dimensions
Vector is levied, wherein M represents the species of view, and N represents the classification number of image.
Step 308.M*N eigenmatrix, stores from per the vector of output characteristic all the way.
Step 309. optimal characteristics matrix, judges optimum by relatively per the size of the characteristic vector of view all the way and regards
Angle, and optimum visual angle vector is marked and is stored.
Preferentially, the comparison of feature sizes is measured using Euclidean distance.
Finally it should be noted that:Above example only to illustrate technical scheme, rather than a limitation;Although
With reference to the foregoing embodiments the present invention has been described in detail, it will be understood by those within the art that:Which still may be used
To modify to the technical scheme described in foregoing embodiments, or equivalent is carried out to which part technical characteristic;
And these modification or replace, do not make appropriate technical solution essence depart from various embodiments of the present invention technical scheme spirit and
Scope.
Claims (9)
1. a kind of multi-view image search method based on deep learning, it is characterised in that comprising training process and retrieving;
Described training process is used for the extraction of characteristics of image in the training of network parameter and image library;
Described retrieving is feature extraction, matching and the output of result for retrieving image.
2. the multi-view image search method based on deep learning according to claim 1, it is characterised in that training process
Including:Multi-view image pretreatment, builds multi views depth convolutional neural networks, and network parameter becomes more meticulous adjustment, in data base
Characteristics of image and optimum visual angle calculate;
Described multi-view image pretreatment, is normalized and standardization processing to being input into multi-view image;
Described structure multi views depth convolutional neural networks, build multichannel depth convolution according to multi views maximum view classification
Neutral net, and using pre-training model parameter as the initiation parameter per multi views convolutional network all the way;
Described network parameter becomes more meticulous adjustment, and the depth convolution per pre-training all the way is adjusted using the multi-view image of labelling
Neural network parameter;
In described data base, characteristics of image and optimum visual angle calculate, and calculate the feature of the image in data base, calculate each group
The optimum multi-view image feature of multi-view image, and these image feature vectors and optimum visual angle are stored.
3. the multi-view image search method based on deep learning according to claim 2, it is characterised in that described is more
View image pretreatment is included graphical rule normalization, image dimension normalization.
4. the multi-view image search method based on deep learning according to claim 2, it is characterised in that the construction
Multi views depth convolutional neural networks are built using VGG-M network parameters, will be replaced with finally by Softmax
newSoftmax。
5. the multi-view image search method based on deep learning according to claim 2, it is characterised in that described number
Calculate according to characteristics of image in storehouse and optimum visual angle, in data base, characteristics of image and optimum visual angle calculate, in calculated off line data base
Multi-view image feature and its optimum visual angle, and result is stored in property data base.
6. the multi-view image search method based on deep learning according to claim 1, it is characterised in that retrieving
Including:Image semantic classification, depth convolutional neural networks feature extraction extract optimum multi-view image, retrieval result sequencing of similarity
Output;
Described Image semantic classification, the image to being input into are normalized and standardization processing;
Described depth convolutional neural networks feature extraction, extracts input picture by training the depth convolutional neural networks for completing
Feature;
Described extraction optimum multi-view image, that is, find out the optimum visual angle of the image of retrieval image similarity;
Described retrieval result sequencing of similarity output, according to sequencing of similarity feedback searching.
7. the multi-view image search method based on deep learning according to claim 6, it is characterised in that the retrieval
As a result sequencing of similarity output, including row distance ratio will be entered between the characteristics of image in characteristics of image to be retrieved and data base
Compared with, sorted according to the numerical value of distance from small to large, and by the optimum multi-view image and similar image itself of these multi-view images
Feedback output.
8. the multi-view image search method based on deep learning according to claim 2, it is characterised in that training process
Comprise the following steps:
Step 1. multi-view image pretreatment, by multi-view image dimension normalization, image dimension normalization, while by multi views
Each view of image makes a distinction classification;Multi-view image data collection is divided into into test data set and training dataset two
Point.
Step 2. constructs multi views depth convolutional neural networks, is that each class view is joined using VGG-M networks according to view classification
Number, simply exports using newSoftmax to replace by Softmax classification by rear, and the network parameter by the use of pre-training is used as net
The initial weight of network.
Step 3. network parameter becomes more meticulous adjustment, reversely adjusts network parameter with training dataset, further with tape label
Test data set come becoming more meticulous adjust network parameter, the multi views deep learning network model after being trained.
In step 4. data base, characteristics of image and optimum visual angle calculate, the feature of the multi-view image in calculated off line data base and
Its optimum visual angle, and result is stored in property data base.
9. the multi-view image search method based on deep learning according to claim 6, it is characterised in that retrieving
Comprise the following steps:
Step one. Image semantic classification, graphical rule normalization that will be to be retrieved, image dimension normalization.
Step 2. by image by any multi views depth convolutional neural networks all the way, it is calculated the feature of image.
Step 3. the feature of image is compared with the characteristics of image in data base, according to distance, output image is indexed from small to large
Number, the corresponding optimum multi-view image of call number is extracted from image library.
Step 4. retrieval result sequencing of similarity is exported, by the optimum multi-view image and similar image packet output of retrieval.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201611014355.1A CN106570192A (en) | 2016-11-18 | 2016-11-18 | Deep learning-based multi-view image retrieval method |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201611014355.1A CN106570192A (en) | 2016-11-18 | 2016-11-18 | Deep learning-based multi-view image retrieval method |
Publications (1)
Publication Number | Publication Date |
---|---|
CN106570192A true CN106570192A (en) | 2017-04-19 |
Family
ID=58542906
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201611014355.1A Pending CN106570192A (en) | 2016-11-18 | 2016-11-18 | Deep learning-based multi-view image retrieval method |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN106570192A (en) |
Cited By (15)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109460492A (en) * | 2018-10-22 | 2019-03-12 | 武汉极意网络科技有限公司 | Method for building up, device, server and the storage medium of GCN model |
CN109522902A (en) * | 2017-09-18 | 2019-03-26 | 微软技术许可有限责任公司 | The extraction of Space-Time character representation |
CN109547410A (en) * | 2018-10-22 | 2019-03-29 | 武汉极意网络科技有限公司 | Request recognition methods, device, server and storage medium based on GCN |
CN109657084A (en) * | 2019-01-07 | 2019-04-19 | 哈尔滨理工大学 | A kind of book retrieval method based on image procossing |
CN109726746A (en) * | 2018-12-20 | 2019-05-07 | 浙江大华技术股份有限公司 | A kind of method and device of template matching |
CN109886191A (en) * | 2019-02-20 | 2019-06-14 | 上海昊沧系统控制技术有限责任公司 | A kind of identification property management reason method and system based on AR |
CN110008999A (en) * | 2019-03-07 | 2019-07-12 | 腾讯科技(深圳)有限公司 | Determination method, apparatus, storage medium and the electronic device of target account number |
CN110020113A (en) * | 2017-09-28 | 2019-07-16 | 南京无界家居科技有限公司 | A kind of family product prediction technique and device based on characteristic matching |
CN110288026A (en) * | 2019-06-27 | 2019-09-27 | 山东浪潮人工智能研究院有限公司 | A kind of image partition method and device practised based on metric relation graphics |
CN110543581A (en) * | 2019-09-09 | 2019-12-06 | 山东省计算中心(国家超级计算济南中心) | Multi-view three-dimensional model retrieval method based on non-local graph convolution network |
CN110555121A (en) * | 2019-08-27 | 2019-12-10 | 清华大学 | Image hash generation method and device based on graph neural network |
CN110852385A (en) * | 2019-11-12 | 2020-02-28 | 北京百度网讯科技有限公司 | Image processing method, device, equipment and storage medium |
CN110992217A (en) * | 2019-11-15 | 2020-04-10 | 广东工业大学 | Method and device for expressing and searching multi-view features of design patent |
CN112989094A (en) * | 2021-02-04 | 2021-06-18 | 北京林业大学 | Specimen information query, sorting and extraction system and method |
CN114020953A (en) * | 2021-10-27 | 2022-02-08 | 北京中知智慧科技有限公司 | Multi-image retrieval method and device for appearance design product |
Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20130201186A1 (en) * | 2012-02-08 | 2013-08-08 | Thomson Licensing | Film grain for stereoscopic or multi-view images |
CN103295025A (en) * | 2013-05-03 | 2013-09-11 | 南京大学 | Automatic selecting method of three-dimensional model optimal view |
-
2016
- 2016-11-18 CN CN201611014355.1A patent/CN106570192A/en active Pending
Patent Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20130201186A1 (en) * | 2012-02-08 | 2013-08-08 | Thomson Licensing | Film grain for stereoscopic or multi-view images |
CN103295025A (en) * | 2013-05-03 | 2013-09-11 | 南京大学 | Automatic selecting method of three-dimensional model optimal view |
Non-Patent Citations (1)
Title |
---|
HANG SU 等: "3D object retrieval based on multi-view convolutional", 《2015 IEEE INTERNATIONAL CONFERENCE ON COMPUTER VISION》 * |
Cited By (21)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109522902A (en) * | 2017-09-18 | 2019-03-26 | 微软技术许可有限责任公司 | The extraction of Space-Time character representation |
CN109522902B (en) * | 2017-09-18 | 2023-07-07 | 微软技术许可有限责任公司 | Extraction of space-time feature representations |
CN110020113B (en) * | 2017-09-28 | 2021-04-20 | 南京无界家居科技有限公司 | Home product prediction method and device based on feature matching |
CN110020113A (en) * | 2017-09-28 | 2019-07-16 | 南京无界家居科技有限公司 | A kind of family product prediction technique and device based on characteristic matching |
CN109547410A (en) * | 2018-10-22 | 2019-03-29 | 武汉极意网络科技有限公司 | Request recognition methods, device, server and storage medium based on GCN |
CN109460492A (en) * | 2018-10-22 | 2019-03-12 | 武汉极意网络科技有限公司 | Method for building up, device, server and the storage medium of GCN model |
CN109726746A (en) * | 2018-12-20 | 2019-05-07 | 浙江大华技术股份有限公司 | A kind of method and device of template matching |
CN109657084A (en) * | 2019-01-07 | 2019-04-19 | 哈尔滨理工大学 | A kind of book retrieval method based on image procossing |
CN109886191A (en) * | 2019-02-20 | 2019-06-14 | 上海昊沧系统控制技术有限责任公司 | A kind of identification property management reason method and system based on AR |
CN110008999A (en) * | 2019-03-07 | 2019-07-12 | 腾讯科技(深圳)有限公司 | Determination method, apparatus, storage medium and the electronic device of target account number |
CN110288026A (en) * | 2019-06-27 | 2019-09-27 | 山东浪潮人工智能研究院有限公司 | A kind of image partition method and device practised based on metric relation graphics |
CN110555121A (en) * | 2019-08-27 | 2019-12-10 | 清华大学 | Image hash generation method and device based on graph neural network |
CN110555121B (en) * | 2019-08-27 | 2022-04-15 | 清华大学 | Image hash generation method and device based on graph neural network |
CN110543581B (en) * | 2019-09-09 | 2023-04-04 | 山东省计算中心(国家超级计算济南中心) | Multi-view three-dimensional model retrieval method based on non-local graph convolution network |
CN110543581A (en) * | 2019-09-09 | 2019-12-06 | 山东省计算中心(国家超级计算济南中心) | Multi-view three-dimensional model retrieval method based on non-local graph convolution network |
CN110852385A (en) * | 2019-11-12 | 2020-02-28 | 北京百度网讯科技有限公司 | Image processing method, device, equipment and storage medium |
CN110852385B (en) * | 2019-11-12 | 2022-07-12 | 北京百度网讯科技有限公司 | Image processing method, device, equipment and storage medium |
CN110992217A (en) * | 2019-11-15 | 2020-04-10 | 广东工业大学 | Method and device for expressing and searching multi-view features of design patent |
CN112989094A (en) * | 2021-02-04 | 2021-06-18 | 北京林业大学 | Specimen information query, sorting and extraction system and method |
CN114020953A (en) * | 2021-10-27 | 2022-02-08 | 北京中知智慧科技有限公司 | Multi-image retrieval method and device for appearance design product |
CN114020953B (en) * | 2021-10-27 | 2022-12-13 | 北京中知智慧科技有限公司 | Multi-image retrieval method and device for appearance design product |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN106570192A (en) | Deep learning-based multi-view image retrieval method | |
CN104834748B (en) | It is a kind of to utilize the image search method based on deep semantic sequence Hash coding | |
CN109685115B (en) | Fine-grained conceptual model with bilinear feature fusion and learning method | |
CN105320965B (en) | Sky based on depth convolutional neural networks composes united hyperspectral image classification method | |
US20190228268A1 (en) | Method and system for cell image segmentation using multi-stage convolutional neural networks | |
CN105205475B (en) | A kind of dynamic gesture identification method | |
CN103544506B (en) | A kind of image classification method and device based on convolutional neural networks | |
CN108021947B (en) | A kind of layering extreme learning machine target identification method of view-based access control model | |
CN106960214A (en) | Object identification method based on image | |
CN108780508A (en) | System and method for normalized image | |
CN107392919B (en) | Adaptive genetic algorithm-based gray threshold acquisition method and image segmentation method | |
CN112036447B (en) | Zero-sample target detection system and learnable semantic and fixed semantic fusion method | |
CN110070107A (en) | Object identification method and device | |
CN109002755B (en) | Age estimation model construction method and estimation method based on face image | |
CN110222215B (en) | Crop pest detection method based on F-SSD-IV3 | |
CN109002562A (en) | A kind of instrument identification model training method and device and instrument recognition methods and device | |
CN107169148A (en) | Image search method, device, equipment and storage medium | |
CN104063713B (en) | A kind of semi-autonomous on-line study method based on random fern grader | |
CN107330750A (en) | A kind of recommended products figure method and device, electronic equipment | |
CN108280488A (en) | Object identification method is captured based on shared neural network | |
CN108133235B (en) | Pedestrian detection method based on neural network multi-scale feature map | |
CN109886342A (en) | Model training method and device based on machine learning | |
CN112507896A (en) | Method for detecting cherry fruits by adopting improved YOLO-V4 model | |
Ghosh et al. | A novel approach to detect and classify fruits using ShuffleNet V2 | |
CN114359199A (en) | Fish counting method, device, equipment and medium based on deep learning |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
WD01 | Invention patent application deemed withdrawn after publication | ||
WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20170419 |